[Insight-developers] HDF5 failing to compile with gcc3.4

[Brad King]

On 05/10/2011 08:40 AM, Gaëtan Lehmann wrote:
>>  #include <stddef.h>
>>
>> in the #ifdef H5_STDC_HEADERS block of H5private.h.
>
> I did try to add that include, without success.

The rest of my questions from the end of that thread will help:

On 04/26/2011 08:59 AM, Brad King wrote:
> That is (indirectly) using CMake's check_type_size() macro which does
>
>     if(HAVE_SYS_TYPES_H)
>       set(headers "${headers}#include <sys/types.h>\n")
>     endif()
>     if(HAVE_STDINT_H)
>       set(headers "${headers}#include <stdint.h>\n")
>     endif()
>     if(HAVE_STDDEF_H)
>       set(headers "${headers}#include <stddef.h>\n")
>     endif()
>
> so that each of those headers is included in the check.  There is an
> option to turn the includes off but HDF5 is not using it.  Do those
> three HAVE_*_H values appear in the cache as ON?  What about
> H5_HAVE_STDINT_H?

What is the state of these cache values?

The H5private.h header also has this:

#ifdef H5_HAVE_UNISTD_H
#   include <sys/types.h>
#   include <unistd.h>
#endif
...
#ifndef __cplusplus
#ifdef H5_HAVE_STDINT_H
#   include <stdint.h>
#endif
#endif

Are either of these being included in the translation units that fail?
(Just add a #error in each block by hand to test it.)

> Brad, I can't give you a direct access to that machine, but maybe we  
> can plan a skype meeting to fix that problem?

Sure, we can do this if we can't resolve it with a couple more emails.

Thanks,
-Brad